一种改进的道路行车密度峰值模糊聚类算法

An Improved Density Peak Fuzzy Clustering Algorithm Based on Millimeter-Wave Radar

为解决城市道路中相邻车辆聚类精度低的问题,本文提出了一种改进的密度峰值模糊聚类算法。首先,该算法使用自适应椭圆距离代替欧式距离,并在决策图中引入指数函数曲线选择密度峰值点,以确定初始聚类中心和聚类数目;接着,将初始信息代入模糊C均值(FCM)聚类算法中,经迭代计算取得一次聚类结果;最后,根据雷达数据中同一辆车的数据点速度差值极小、不同车辆的速度差值相对较大这一特征,引入和速度相关的目标函数,并通过迭代计算取得最终的聚类结果,以对一次聚类结果进行修正。根据真实道路测量数据的实验证明,本文提出的聚类算法精度高、鲁棒性好,能正确聚类城市道路中相邻的车辆目标,具有更好的聚类效果。为道路中车辆的跟踪、交通状态预估等处理提供可靠、准确的目标信息,大大减少后续工程的计算量。

In order to solve the problem of low clustering accuracy of adjacent vehicles on urban roads,an improved density peak fuzzy clustering algorithm is proposed in this paper.Firstly,the algorithm uses adaptive ellipse distance instead of Euclidean distance,and introduces an exponential function curve in the decision diagram to select density peak points to determine the initial cluster center and number of clusters.Secondly,the initial information is substituted into the Fuzzy C-means(FCM)clustering algorithm,and first clustering result is obtained through iterative calculation.Finally,according to such characteristics as extremely small speed difference of the data points of the same car in the radar data and relatively large speed difference of different vehicles,an objective function related to the speed is introduced.The final clustering result is obtained through iterative calculation to correct the first clustering result.The experiments with real road measurement data show that the proposed clustering algorithm has high accuracy and good robustness and can correctly cluster adjacent vehicle targets on urban roads with better clustering effect.The method can provide reliable and accurate target information for the vehicle tracking and traffic state estimation,greatly reducing the amount of calculation in the subsequent projects.